1. Introduction

Voice pitch is the most salient acoustic feature of human speech (Banse & Scherer, 1996). Voice pitch is also sexually dimorphic; males speak up to an octave lower than females do, on average (e.g., Klatt & Klatt, 1990). This sexual dimorphism is not due merely to sex differences in body size. Relative to both height and body volume, vocal fundamental frequency (F0, the acoustic parameter most closely related to pitch) is lower in adult males than in adult females and prepubescent children of both sexes (Titze, 2000).

The proximate and developmental causes of this male secondary sex characteristic are known. Adult males have 60% longer membranous portions of the vocal folds, which vibrate at a lower F0. Vocal anatomy is masculinized under the influence of pubertal androgens, which cause males' vocal folds to increase in size faster than the overall rate of body growth (Hollien, 1960). Moreover, circulating androgens appear to maintain masculine vocal fold morphology in adults; F0 correlates negatively with androgen levels in men (Dabbs & Mallinger, 1999) and decreases with androgen treatment (Need, Durbridge, & Nordin, 1993).

By contrast, little is known about why men evolved this developmental pattern. Some evidence suggests that low F0 increased males' attractiveness to females. Oguchi and Kikuchi (1997) found that male voices rated as highly attractive by Japanese university students were significantly lower in voice pitch (VP) than were the voices rated as less attractive. Additionally, Collins (2000) found that Dutch women rated lower male voices as being significantly more attractive. However, because of their correlational designs, these studies could not demonstrate that pitch (P), in itself, affects vocal attractiveness.

Nonetheless, if female choice shaped male VP over human evolution, this preference requires explanation. Mate preferences are costly and generally evolve because they provide compensatory benefits (Andersson, 1994). In many species, female preferences apparently function in recruiting mates of high genetic quality, and androgen-dependent traits are often the foci of such preferences (Andersson, 1994). The degree of expression of androgen-dependent traits may signal fitness because androgens are immunosuppressant (Grossman, 1985) and can be produced at high levels only by otherwise healthy individuals (Folstad & Karter, 1992; see also Zahavi & Zahavi, 1997). Evidence from nonhuman animals (reviewed in Andersson, 1994) and humans (e.g., Gangestad & Thornhill, 2003, Scheib et al., 2003) suggests that some masculine traits are reliable fitness indicators, and genetic transmission of these fitness benefits has been demonstrated in some species (e.g., Norris, 1993, Reynolds & Gross, 1992).

In humans, female preferences for androgen-dependent (masculine) features show signs of complex design (Williams, 1966) for recruiting good genes. For example, women's mate preferences vary with probability of conception, such that women in the follicular (fertile) phase of their menstrual cycles exhibit preference shifts toward more masculine facial features (Johnston et al., 2001, Penton-Voak et al., 1999, Penton-Voak & Perrett, 2000) and odors (Grammer, 1993). Moreover, women's preferences for androgen-dependent features may be greater when evaluating men as short-term sex partners rather than as long-term mates (Penton-Voak et al., 1999; reviewed in Penton-Voak & Perrett, 2001). Because low VP is androgen dependent, it may also signal heritable fitness, and women's preferences for low VP may similarly depend on mating context and menstrual phase. Indeed, some studies have found that women prefer the voices of men who carry other fitness markers, such as external body symmetry and a high shoulder-to-hip ratio (Hughes et al., 2004, Hughes et al., 2002).

The present study thus evaluates the predictions of one hypothesis for the evolution of low voice pitch in human males: Women prefer males with low voices because VP reflected heritable fitness in ancestral environments. Specifically, I test whether low male VP (1) increases short-term, sexual attractiveness more than long-term relationship attractiveness, (2) is more attractive to women nearer peak fertility, and (3) predicts higher mating success. Unlike previous studies (Collins, 2000, Oguchi & Kikuchi, 1997), this study uses an experimental, rather than correlational, design by examining the effects of computer-based pitch manipulations on attractiveness ratings.

2. Methods

3. Results

An initial regression analysis revealed a moderate but significant linear relationship—and, in each case, a less significant quadratic relationship—between unmanipulated VP and each of the four categories of attractiveness ratings (short-term/fertile, short-term/nonfertile, long-term/fertile, and long-term/nonfertile). Linear relationships between F0 (predictor variable) and attractiveness ratings ranged from p=.003, β=−.27, and r2=.07 for short-term/fertile to p=.036, β=−.17, and r2=.03 for long-term/nonfertile (p=.023, β=−.22, and r2=.05 for short-term/nonfertile; p=.006, β=−.26, and r2=.07 for long-term/fertile; n=111 male voices for all regressions). Thus, at least over the normal range, male VP appears to be linearly and negatively associated with attractiveness ratings.

When the voices experimentally raised by one semitone were compared with the same voices lowered by one semitone under each of the four mating context-by-fertility conditions, multifactor repeated-measures analysis of variance (ANOVA) revealed no significant main effect of P manipulation [F(1,106)=1.31, n.s.]. On the other hand, there was a significant three-way interaction between P manipulation, mating context, and fertility [F(1,106)=3.24, p=.037], as predicted (Fig. 1). Lowered-P recordings obtained significantly higher attractiveness scores than did the raised-P recordings only for short-term/fertile attractiveness [paired t(106)=2.09, p=.020]. Equivalent comparisons showed that lowered- and raised-P recordings did not differ significantly in long-term/fertile [t(106)=0.37, n.s.], long-term/nonfertile [t(106)=0.47, n.s.], or short-term/nonfertile [t(109)=1.11, n.s.] attractiveness scores. A significant two-way interaction was also found between P manipulation and mating context [F(1,106)=8.61, p=.002] but not between P manipulation and fertility [F(1,106)=0.88, n.s.].

Fig. 1. Three-way interaction between P manipulation, relationship context, and fertility [F(1,106)=3.24, p=.037]. VP manipulation significantly affected attractiveness ratings only for women in the fertile phase of their cycles evaluating men for short-term, sexual relationships [paired t(106)=2.09, p=.020].

Finally, unmanipulated VP (measured by F0) negatively predicted male participants' self-reported number of sexual partners over the past year [Pearson's r(103)=−.17, p=.043].

4. Discussion

Women prefer the utterances of men who speak at a low pitch. This conclusion is supported by previous studies (Collins, 2000, Oguchi & Kikuchi, 1997) as well as by the present study, in which significant negative linear relationships were observed between the F0 of spontaneous male utterances and women's ratings of those utterances. However, these correlations do not appear to be driven by an effect of VP itself on attractiveness ratings. In the present study, variations among utterances, such as those due to content, tempo, amplitude, enunciation, and accent, were experimentally controlled by manipulating VP and comparing utterances to themselves at a different P. The result was that VP had no main effect on attractiveness ratings.

Although experimental P manipulation did not influence attractiveness to females as a main effect, it did affect attractiveness under specific conditions. This was demonstrated by a significant three-way interaction between P manipulation, mating context, and fertility status. Lowered-P voices were significantly more attractive than the same voices raised in P, but only to women in the fertile phase of their ovulatory cycles, rating the men's voices with respect to a short-term mating context. Attractiveness ratings of lowered-P and raised-P voices did not differ in any other mating context-by-fertility status condition.

A significant two-way interaction was also found between VP manipulation and mating context. An order effect of rating short-term attractiveness before long-term attractiveness seems unlikely to explain this result because VP would have had to affect ratings differently depending on rating order. More probably, VP has a greater effect on sexual attractiveness than it does on attractiveness for committed relationships. The increased importance of VP for short-term mating contexts may help explain why men who spoke at a lower F0 reported more sexual partners in the past year.

The above evidence is consistent with the hypothesis that women prefer fertile-phase sexual liaisons with low-voice-pitch males because such behavior acquired favorable genes for the offspring of ancestral females. Males who are putatively of high genetic quality (those with high testosterone levels) appear to invest less in their mates (reviewed in Penton-Voak & Perrett, 2001, Puts, 2003). It may, therefore, have benefited ancestral females to prefer such males primarily for sexual, rather than committed, relationships and mainly during the fertile menstrual phase.

The present study thus supports the role of female mate choice for good genes in shaping male voice pitch. However, alternative interpretations are possible. For example, low voices may have helped ancestral males win mates through male dominance competition (Ohala, 1982, Ohala, 1983, Ohala, 1984, Puts, in review). This explanation does not account for the observed menstrual- and mating context-related patterns in women's preferences for male voice pitch, but the two explanations are not mutually exclusive.

Male voice pitch and female preferences for it could also have evolved via Fisherian “runaway sexual selection” (Fisher, 1930). According to this model, the heritable benefit associated with a male trait is increased attractiveness of sons. Although a widespread sexual preference of fertile-phase women could plausibly maintain low male VP in human populations, this hypothesis does not explain how the female preference became widespread in the population. According to Fisher (1930), a female preference initially spreads because the preferred male trait is linked to some other selective benefit. As suggested above, masculine traits such as low VP may be linked to heritable health (see Section 1). Additionally, some evidence indicates an association between fertilization ability and the expression of testosterone-dependent traits such as low VP in nonhuman animals (Andersson, 1994) and humans (Puts, in press). For example, in the Australian frog Uperoleia laevigata, females prefer larger males, which produce lower frequency calls and leave fewer eggs unfertilized (Robertson, 1990). An association between fertilization ability and VP in men could also select for mating context- and fertility status-related variation in women's preferences for male VP (Puts, in press).

In summary, women in the fertile phase of their menstrual cycles prefer low male voice pitch in short-term mating contexts, and males who speak at lower pitches when courting females report more sexual partners. Female preferences are thus likely to have been important selection pressures on male voice pitch and probably contributed to human sexual dimorphism in this trait, although male dominance competition may also have been involved. Female preferences for low male VP may have evolved because males with a low VP offer increased fertilization ability or heritable benefits, including healthier and more attractive offspring.